A Novel Approach for Face Recognition: YOLO-Based Face Detection and Facenet

Advances in information technology encourage digitization of all aspects of human life, including government. The Indonesian government also uses the Electronic Based Government System (SPBE) to transform governance by utilizing information and communication technology to provide services to SPBE users. However, with this transformation, there are weaknesses and loopholes that irresponsible people, such as attacks on systems, theft of identity and personal data, and falsification of electronic documents, can exploit. Therefore, a system is needed to verify the user's identity. To meet these needs, we developed a face recognition system that uses faces to recognize and verify a person's identity. We develop a face recognition system by utilizing an existing pre-trained model, the YOLO (You Only Look Once) deep learning model, as a face detection system with various weights and different versions. Furthermore, the face detection system will be integrated with Facenet to perform feature extraction and face classification. The system will be tested and rated based on the average time it takes to process images and their accuracy. After testing the dataset consisting of 100 identities and 200 student facial images, the best results obtained are the YOLOv5 model and yolov5s weights with a processing time of 183 ms and an accuracy of 99.5 %.